Errors in variables and/or total least squares

["Hunsicker, Lawrence" <lawrence-hunsicker@uiowa.edu>]

Good morning folks:

I have an errors in variables problem that I would like to have some help on.  There are several ways to estimate an underlying unobserved variable.  The various methods can be assumed to give results that are linearly related to the underlying latent variable, but each estimates the latent variable with error.  It is a reasonable initial assumption that these various errors are uncorrelated, since the imprecisions relate to different inaccuracies in measurement.  I would like to know: 

        a)  Do these various methods actually measure the same underlying latent variable?

        b)  If the answer to a) is “yes,” which of these various methods is most precise, and therefore the best one to use to estimate the latent variable, and how much better?

Are there any S-Plus packages that deal well with this kind of question?  One specific method is Total Mean Squares.  Is there a function that measures this?

Many thanks in advance for any help that you can give me.

Larry Hunsicker